Which Tape Drive blocksize should I choose?

[Kazuha]

Hi everyone!
The default tape drive blocksize in Veeam is 256K(262144), although there's a bigger blocksize option in Veeam.

But many people say the performance will be better as the blocksize gets bigger, is that true?
If it is true, why not Veeam uses 512K(524288) or bigger blocksize as default to get better performance?
And what's the difference between different blocksize?

[HannesK]

Hello,

But many people say the performance will be better as the blocksize gets bigger

do you maybe have a reference (link) for that? What did you see that in your environment?

The block size can be set manually, in case auto-detection does not work.

Best regards,
Hannes

[Kazuha]

Hi Hannes! Nice to meet you again!

In https://forums.freebsd.org/threads/back ... ost-152533

@AndyUKG: For LTO drives normally you want a large block size, at least 64K but for performance more optimatally 256K or larger.

In https://www.ibm.com/docs/en/zos/2.1.0?t ... r-than-32k

A larger tape block size can improve elapsed time and tape utilization. The larger block sizes allow the tape device to perform better because of the decreased need to start and stop reading blocks.

In https://www.veritas.com/support/en_US/article.100019307

This larger block size may or may not have a performance benefit to operations.

But some of these links are about 10ys ago, at the age of 32 or 64kb blocksize were commonly used. So I don't know whether the theory is still valid nowadays.

And about a year ago, someone asked a similar question tape-f29/lto8-and-blocksizes-t74657.html and Dima said:

Hm, if I remember correctly the highest possible value is reported to Veeam B&R by the tape device itself (and is used automatically) but we left the selection available for old devices that do not have this logic.

But I found my tape drive can support 524288(512K) blocksize but Veeam set it to 262144(256K) as default.

And I found my tape drive speed never exceed 140M/s even if I enabled compression.
The Wikipedia https://en.wikipedia.org/wiki/Linear_Tape-Open says lto5 can reach 280M/s when the data is very compressible.
So here's another question: Where can I see the compression ratio of each tape? I can only see the free space for each tape in Veeam.

Best regards
Kazuha

[HannesK]

Hello,
yes, I also found these old posts In your environment, the value is set to 256KB. That is the value recommended in some of the old posts.

And I found my tape drive speed never exceed 140M/s even if I enabled compression.

that's the physical limit. Compression on the tape device is irrelevant. Veeam backups are compressed already. "Double compression" is waste of resources.

Looks like everything is working at maximum physical speed in your environment.

Best regards,
Hannes

[Kazuha]

Hi Hannes!
So the speed in Veeam log GUI is compressed physical speed?

And if I set blocksize to 512kb, will it do any benefit to my backup progress(eg.faster seeking or faster writing)?
Sincerely,
Kazuha

[HannesK]

Hello,
the speed shown in the console is the physical speed, yes.

will it do any benefit to my backup progress(eg.faster seeking or faster writing)?

as I cannot test myself in your environment, I cannot answer that question. I doubt, that it has impact (because you are already at the physical limit) and that's the reason why I asked in my first answer what differences in speed you saw

Best regards,
Hannes

[Kazuha]

Hi Hannes!

Wow, so it seems enough to set 256k blocksize, thanks Hannes!

Cheers!
Kazuha

[Dima P.]

Kazuha,

We set the maximum possible value received from the hardware be default. I'd say it's ok to keep it as-is however you may need to change this value whenever tape was recorded with one drive/lower block size and then you perform restore with another drive/different block size. Usually it happens when tape infrastructure is migrated or upgraded.

[Kazuha]

Hi Dima! Nice to meet you!

So my drive(HP LTO5) can support 512K blocksize max(I can choose 512K in the Veeam console but I haven't tested if 512K blocksize can work),
why does Veeam set the blocksize to 256K by default?
Just wondering, I think 256K is also a good blocksize

Sincerely,
Kazuha

[RGijsen]

Sidequestion regarding this; we don't use tape ourselves, but we have ran into block-size issues in the past with regards to copy jobs. In our main site we wanted a large block size, and in our off-site backup we wanted to use a smaller block size as that side works with ReFS block cloning and also the built-in deduplication works better with smaller blocksize. However, one can't change the block-size per repository, in the sense that for a copy job, source and destination MUST use the same blocksize.

Just to learn, is that not applicable with cope-to-tape jobs?

[Dima P.]

Kazuha,
Confirmed that 256k is actually default block size we use for all the drives (that's why it's set as default). You can use 512k for your drive without any issues.

RGijsen,
This thread is unrelated to the block size in the backup / backup copy jobs, but since you've asked - backup copy uses block size from source backup file and there is not way to change it.

[JaySt]

Coming back to this: to confirm, does veeam set the default dynamically or is the default statically set to 256K ? Not sure i got the conclusion right after reading above posts

[Dima P.]

Hi Jay,

256k is the default block size set for all the drives. Thanks!